Decentralized Concensus Discussed
Vitalik Buterins claim that to achieve, higher transactions per second you need to be “centralized” in the sense to you are running few nodes, is not really correct, it assumes a certain amount of things.
A decentralized consensus system where all Nodes connected is a part of the consensus has two problems, it do not scale, as the more nodes you add in, the more total amount of data will need to be moved in a specific timeframe over an unknown network, in the sense that you will not know where globally the Node is deployed, secondly a consensus based on all Nodes are a part of the consensus is vulnerable to 51% attack as there is no intelligence in the consensus a brute force attack can take it over.
So how could a secure consensus be achieved with a smaller group of nodes and still be decentralized, if we interpret decentralized as no Node or group of Nodes can ever work independent of the mass, then only a brute force consensus is available to us and the 51% attack will always exist.
But, if we accept that moving data over the Internet is the base of the consensus then things get interesting, as the Internet is not homogenous the idea with a all-for-one consensus is really not that smart, the Internet has a global “data state” and multiple local “data states” as there are caches, buffering, proxy’s, to help speed up data transactions, we can see that working with local and global states for consensus will make sense, and by using a pseudo random selection of signing nodes for each local “data state” you can achieve area based micro signing that is then signed in to the main chain.
So to exemplify, transactions done in New York can be signed in New York and will not need to be distributed globally to get signed. To wrap up.
- Create a higher transaction throughput of transactions by “micro signing” at local network areas.
- Global verifications of blocks not transactions.
- No One-For-All consensus as it opens up for 51% attack.